popov/neuron-tai
1
0
Fork 0
Code Issues 42 Pull Requests Actions Packages Projects Releases Wiki Activity

DGR-051: Assemble the DeepSeek V4 Flash ShardEngine adapter #35

New Issue
Open
opened 2026-07-16 20:36:05 +00:00 by popov · 0 comments
popov commented 2026-07-16 20:36:05 +00:00
Owner
Copy Link

DGR-051: Assemble the DeepSeek V4 Flash ShardEngine adapter

  • Status / triage: specification only; ready-for-agent; passes: false
  • Execution mode: AFK
  • Milestone: M3
  • Dependencies: DGR-034, DGR-037, DGR-047, DGR-048, DGR-049, DGR-050
  • Blocks (derived): DGR-052, DGR-060, DGR-065, DGR-069
  • Labels: area:deepseek-v4, area:engine, type:vertical-slice, priority:p0, ready-for-agent
  • Evidence class: real-model
  • Hardware: optional
  • Model: deepseek-v4-flash
  • Upstream: yes

Objective / description

Fresh Ralph session: read .scratch/distributed-gguf-runtime/RALPH-CONTEXT.md, source issue .scratch/distributed-gguf-runtime/issues/051-assemble-the-deepseek-v4-flash-shardengine-adapter.md, and evidence READMEs for dependencies (DGR-034, DGR-037, DGR-047, DGR-048, DGR-049, DGR-050) before changing code. Inspect live source/tests rather than trusting legacy pass states. Objective: Assemble range ownership hooks and boundary/local-state adapters around upstream llama.cpp V4 execution behind the stable ShardEngine interface.

Acceptance criteria

  • Support valid head, middle, tail, and whole-model ranges over all 43 main layers using the pinned upstream llama.cpp V4 graph/kernels.
  • Integrate exact range ownership, mHC boundary, positions, required token-ID sideband, and shard-local session/epoch state adapters.
  • Report MTP reserved-but-not-executable for alpha and reject unsupported split points or cache expectations.
  • Include exact adapter/schema/runtime/artifact identity and valid split restrictions without reimplementing V4 MoE, attention, mHC, or hash routing.
  • Applicable shared quality gates in prd.json pass, and the evidence handoff records exact commands/results, changed files, limitations, and dependency handoff.

Shared quality gates

  • Targeted deterministic tests pass; Python changes also pass python -m compileall packages tests.
  • git diff --check passes.
  • Default tests are model-download-free, API-credit-free, and GPU-free.
  • Evidence README records exact changed files, commands/results, limitations, and dependency handoff; no fabricated evidence or inherited completion credit.

Evidence handoff

Write and verify .scratch/distributed-gguf-runtime/evidence/DGR-051/README.md. Until every criterion and applicable gate has real evidence, this story remains passes: false. Legacy evidence is provenance only, not completion credit.

<!-- GENERATED FROM prd.json — DO NOT EDIT AS AN INDEPENDENT SOURCE. prd.json IS AUTHORITATIVE. --> # DGR-051: Assemble the DeepSeek V4 Flash `ShardEngine` adapter - **Status / triage:** specification only; `ready-for-agent`; `passes: false` - **Execution mode:** `AFK` - **Milestone:** `M3` - **Dependencies:** `DGR-034`, `DGR-037`, `DGR-047`, `DGR-048`, `DGR-049`, `DGR-050` - **Blocks (derived):** `DGR-052`, `DGR-060`, `DGR-065`, `DGR-069` - **Labels:** `area:deepseek-v4`, `area:engine`, `type:vertical-slice`, `priority:p0`, `ready-for-agent` - **Evidence class:** `real-model` - **Hardware:** `optional` - **Model:** `deepseek-v4-flash` - **Upstream:** `yes` ## Objective / description Fresh Ralph session: read `.scratch/distributed-gguf-runtime/RALPH-CONTEXT.md`, source issue `.scratch/distributed-gguf-runtime/issues/051-assemble-the-deepseek-v4-flash-shardengine-adapter.md`, and evidence READMEs for dependencies (DGR-034, DGR-037, DGR-047, DGR-048, DGR-049, DGR-050) before changing code. Inspect live source/tests rather than trusting legacy pass states. Objective: Assemble range ownership hooks and boundary/local-state adapters around upstream llama.cpp V4 execution behind the stable ShardEngine interface. ## Acceptance criteria - [ ] Support valid head, middle, tail, and whole-model ranges over all 43 main layers using the pinned upstream llama.cpp V4 graph/kernels. - [ ] Integrate exact range ownership, mHC boundary, positions, required token-ID sideband, and shard-local session/epoch state adapters. - [ ] Report MTP reserved-but-not-executable for alpha and reject unsupported split points or cache expectations. - [ ] Include exact adapter/schema/runtime/artifact identity and valid split restrictions without reimplementing V4 MoE, attention, mHC, or hash routing. - [ ] Applicable shared quality gates in `prd.json` pass, and the evidence handoff records exact commands/results, changed files, limitations, and dependency handoff. ## Shared quality gates - Targeted deterministic tests pass; Python changes also pass `python -m compileall packages tests`. - `git diff --check` passes. - Default tests are model-download-free, API-credit-free, and GPU-free. - Evidence README records exact changed files, commands/results, limitations, and dependency handoff; no fabricated evidence or inherited completion credit. ## Evidence handoff Write and verify `.scratch/distributed-gguf-runtime/evidence/DGR-051/README.md`. Until every criterion and applicable gate has real evidence, this story remains `passes: false`. Legacy evidence is provenance only, not completion credit. <!-- ralph-story-id: DGR-051 -->
popov added this to the M3 milestone 2026-07-16 20:36:05 +00:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
area:admission
area:artifacts
area:attention
area:backpressure
area:batching
area:boundary
area:build
area:certification
area:cleanup
area:compression
area:concurrency
area:deepseek-v4
area:engine
area:failure
area:gitea
area:grpc
area:hardware
area:identity
area:integration
area:kv-cache
area:lifecycle
area:llama
area:long-context
area:maintenance
area:meshnet
area:mhc
area:moe
area:mtp
area:node
area:packaging
area:parity
area:patches
area:performance
area:planning
area:protocol
area:provenance
area:recovery
area:relay
area:release
area:reliability
area:routing
area:routing-state
area:scale
area:scheduler
area:supervision
area:testing
area:tracker
area:upstream
area:weights
area:worker
gate:hitl
milestone:alpha
milestone:beta
priority:p0
priority:p1
priority:p2
ready-for-agent
ready-for-human
status:blocked
Blocked by incomplete dependencies or human gate
 status:completed
Completed and verified in prd.json
 status:in-progress
Currently selected by Ralph
 status:ready
Ready for Ralph execution
 type:architecture
type:audit
type:benchmark
type:certification
type:collaboration
type:contract
type:delivery
type:domain
type:fixture
type:hardening
type:infrastructure
type:integration
type:operations
type:optimization
type:performance
type:protocol
type:provenance
type:research-spike
type:runtime
type:simulation
type:toolchain
type:tooling
type:vertical-slice
No Label area:deepseek-v4 area:engine priority:p0 ready-for-agent status:blocked type:vertical-slice
Milestone
No items
No Milestone M3
Projects
Clear projects
No project
Assignees
Clear assignees
deploy popov
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: popov/neuron-tai#35
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?